Problem with #doctest: +FLOAT_CMP on Windows Py 2.6 ?

PR #3731 at ed8a3e1 failed only on Windows Py 2.6 with an exception in doc testing. There were no other test failures on any platform / configuration. I don't have a Windows platform to debug or try to reproduce.

Here is the log:

https://ci.appveyor.com/project/Astropy/astropy/build/1.0.1509/job/65hvtqa4x4h7m3b6

And the error at the end:

================================== FAILURES =================================== 
_________________________ [doctest] access_table.rst __________________________ 
_pytest.runner:137: in __init__ 
>   ??? 
_pytest.runner:124: in <lambda> 
>   ??? 
_pytest.main:161: in call_matching_hooks 
>   ??? 
_pytest.core:380: in pcall 
>   ??? 
_pytest.core:387: in _docall 
>   ??? 
_pytest.core:288: in execute 
>   ??? 
_pytest.runner:86: in pytest_runtest_call 
>   ??? 
astropy\tests\pytest_plugins.py:156: in runtest 
>           raise_on_error=True, verbose=False, encoding='utf-8') 
C:\conda\envs\test\lib\doctest.py:1973: in testfile 
>       runner.run(test) 
C:\conda\envs\test\lib\doctest.py:1736: in run 
>       r = DocTestRunner.run(self, test, compileflags, out, False) 
C:\conda\envs\test\lib\doctest.py:1391: in run 
>           return self.__run(test, compileflags, out) 
C:\conda\envs\test\lib\doctest.py:1269: in __run 
>               if check(example.want, got, self.optionflags): 
astropy\tests\output_checker.py:154: in check_output 
>           return self.normalize_floats(want, got, flags) 
astropy\tests\output_checker.py:117: in normalize_floats 
>           got = got % tuple(nw_) 
E           TypeError: a float is required 
 1 failed, 9031 passed, 111 skipped, 78 xfailed, 2 xpassed in 1118.24 seconds = 
Command exited with code 1

Skip some doc tests to avoid issue astropy#3853

Weird--could you point me to the specific test that's failing? It could be a quirk in the platform float parsing. The value in the nw_ variable should be a float string.

@embray

@embray - unfortunately I can only narrow it down to one of three tests. As you can see in 8c850ef I just skipped all three of the FLOAT_CMP tests and it then passed on windows. I don't have a Windows platform to try reproducing and/or skipping one at a time.

If called as a function then one can supply an ``option`` that specifies
the type of information to return.  The built-in ``option`` choices are
``attributes`` (column attributes, which is the default) or ``stats``
(basic column statistics).  The ``option`` argument can also be a list
of available options::

  >>> t.info('stats')  # doctest: +FLOAT_CMP
  <Table length=5>
  name mean      std      min max
  ---- ---- ------------- --- ---
     a  6.0 4.24264068712   0  12
     b  7.0 4.24264068712   1  13
     c  8.0 4.24264068712   2  14

  >>> t.info(['attributes', 'stats'])  # doctest: +FLOAT_CMP
  <Table length=5>
  name dtype   unit   format       description        mean      std      min max
  ---- ----- -------- ------ ------------------------ ---- ------------- --- ---
     a int32 m sec^-1  %6.3f unladen swallow velocity  6.0 4.24264068712   0  12
     b int32                                           7.0 4.24264068712   1  13
     c int32                                           8.0 4.24264068712   2  14

Columns also have an ``info`` property that has the behavior and arguments,
but provides information about a single column::

  >>> t['a'].info
  name = a
  dtype = int32
  unit = m sec^-1
  format = %6.3f
  description = unladen swallow velocity
  class = Column
  n_bad = 0
  length = 5

  >>> t['a'].info('stats')  # doctest: +FLOAT_CMP
  name = a
  mean = 6.0
  std = 4.24264068712
  min = 0
  max = 12
  n_bad = 0
  length = 5

That's fine, I should be able to get to the bottom of this.

Skip some doc tests to avoid issue astropy#3853

@taldcroft

@taldcroft - do you know if this is a problem on Python 2.7 too? Since we don't support Python 2.6, maybe this can be closed?

@taldcroft

This wouldn't be an issue on Python 2.7. However, before closing this the tests that @taldcroft marked skip in 2cfaf55 can be re-enabled.

@embray

@embray @astrofrog - #4594 will test whether this is now working in our current test suite.

Since we're dropping Python 2.6 support anyways this can be closed.

Actually #4594 hasn't been merged yet so this should be left open I guess. Any reason for the delay on that?

@embray

@embray - yes, #4594 is failing on AppVeyor, meaning that the fix was likely needed for other Python versions too, or that at least something is not quite right in that PR.

@taldcroft

@taldcroft - can this be closed?

Navigation Menu

Repository navigation

Problem with #doctest: +FLOAT_CMP on Windows Py 2.6 ? #3853

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Problem with #doctest: +FLOAT_CMP on Windows Py 2.6 ? #3853

Description

Activity

embray commented on Jun 15, 2015

taldcroft commented on Jun 15, 2015

embray commented on Jun 15, 2015

astrofrog commented on Jan 29, 2016

embray commented on Jan 29, 2016

taldcroft commented on Feb 10, 2016

embray commented on May 23, 2016

embray commented on May 23, 2016

astrofrog commented on May 23, 2016

astrofrog commented on Jan 16, 2017

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions